Testing the dynamic performance of an OPT requires an accurate model.
Interpreting the original Williamson OPT specifications , yields the above
HF and LF models where Lp should be greater than 95H and Ll less than 35mH for 10k p-p KT66 drive

Notable is that the above suggested HF model will not always fit. For some OPT's it is actually
more accurate to use the LF model whith the Lp changed to an interwinding capacitance, Ci,

Except the above obvious and necessary dynamic requirements, the OPT will also have
to have the following important static and time-domain characteristics:

1) High incremental inductance, Li, at low driving voltages (100H@5V/50Hz typ)
2) Higher than or equal to Li at highest driving voltage AND lowest frequency (>230V/20Hz typ)
3) Higher than 50% of Li at modest dc (read unbalance) currents (>10mA@Np/2 typ)
4) A primary dc resistance, Rdc(prim), less than 2.5% of RL (<250Ohm typ)
5) A total freedom from HF oscillations

My first (obsolete) OPT prototype, the KE57613-A, was actually wounded on standard "hard" iron.
Testing the OPT with regard to the dynamic parameters made my hair rise (Ll=28mH, Lp=470H!).
At first impression this seamed to be a perfect choice. But sadly enough the L(dI) (Lp/3@7mA!)
performance was too bad. I tried to suggest air-gaps but that was impossible for my manufacturer.

The solution was my KE57613-C/D OPT wounded on "soft" iron instead (Ll=47mH, Lp=80H). While the dynamic
performance of this transformer not by far reaches my first prototype, the obviously important
(my first Williamson amp actually oscillated at ULF while being driven by LF transients) L(dI)
performance was by far the better (Lp/2@10mA). Moreover the inductance almost stops dropping
for unbalance currents greater than 30mA (@25H) while the inductance almost vanished
(i.e core saturation) at unbalance currents greater than 10mA for the "hard" iron
(see Primary inductance vs DC unbalance for my universal transformer 87060).

With my KE57613-C/D ALL critical parameters above are satisfied (with some small amount
of "phase correction"-feedback and output boot-strapping anyway).

Generally a lower drive/load impedance will improve bass response but worsen treable. Vice
versa is of cource true for higher drive/load impedances. This is however only true as long as
the HF model in fig.3 above holds.


Remove the secondary load and use the secondary winding of "the other" OPT (U0=6.5V).
Connect a 1kOhm resistor i series for easy and quick current calcuation.
Divide U(Lp) or TPC (see fig.1) with I which yields XL. wLp=XL

wl*Lp=2rp//Zla-a, where Zla-a=SQR(n)*RL (3.3Hz@Williamson)

2) TESTING Lleak
Connect according to fig.1 above. Use a at least 20kHz generator, G. Use 5V (doesn't
really matter at these high frequencies, but for power output compatibility)
Measure phase difference, Phi, between TPA and TPB (at the secondary, obs!).

It has been suggested a very simple (but often inaccurate) procedure in testing Lleak.
The method I'm thinking of is "connect the primary windings in serial antiphase and measure
in the same way as Lp and substract Rdc". It can been shown that this method ONLY suffice when
both (or all) primary windings are of EXACTLY equal length AND Rdc (not always a function of
actual length...) AND inductance. A more accurate and universal true method is "the hard way"
already described. As a "Worst case" estimate, this simple method however suffice. Remember
though that Ll might be several magnitudes better/lower.

Measuring Ll for my obsolete OPT in this simple manner, yielded a result of almost 100mH while
the always accurate "hard-way"-method yielded a remarkable Ll of 28mH only.

Some people might say "that transformer is badly balanced" but there are several resons
why this statement may be false. A few is askewed winding and/or torged winding which both
yields the same number of turnes but a less inductance and/or higher resistance respectively.
Anyway, leakage inductance (i.e HF response) is NOT the only result of this simple test.

wh*Ll=(Zla-a+2rp) (60kHz@Williamson)

3) TESTING Rdc...!

Connect primary windings in serial antiphase and measure (use of a high-impedance DVM will
otherwise cause reading oscillations due to high inductance!)

This test is preferably done on the secondary winding whereafter (the low!) voltages and inductances
are transformed to the primary.

You may again use "the other" OPT's secondary winding. If a 230V mains outlet (or KE57613-D)
is available you may also be able to load the DUT transformer with approximately twice it's normal
secondary voltage (2*6.5V=13V). If the mains frequency is 50Hz, the SQR(n)*Ls(DUT) evaluation is equivalent to Lp@230V/25Hz
and this inductance should still and definitely be greater than the incremental inductance measured in point 1
above (95H for a Williamson OPT).

See www.freeyellow.com/members5/rogerk/page9.html for more info/hints.

See point 4 above. Switch DUT testing side to the primary winding. Add a 1.5V battery and a 100Ohm resistor
in series. Measure the ac and dc current as well as ac and dc voltage over the whole DUT
primary. The dc voltage should be around 1V. The ac voltage will be around 6.5V.
Double the dc current for dI(Np/2). Calculate L(dI) which should be greater than Lp/2 (or 50H in my case).

The normal anode voltage difference between mathed tubes while using my OPT's (Rdc=200Ohm) are
often in the order of 0.5-1Vdc, which generally means an anode difference current of 5-10mA.
Generally the tubes should always have a lower anode difference current than 5mA
(or 0.5V in my case). If not, it is strongly recommended to switch tubes or in emergency cases,
use something like the original static-balancing Williamson method.

See www.freeyellow.com/members5/rogerk/page10.html for more info/hints.

This time-domain test is often forgotten. If not tested or considered it may show up as
a persistant output HF oscillation when the (even open-looped) amplifier is driven by HF
transients. This is specially true when feedback is being used. Actually, feeding the output
signal back from the secondary of an OPT with HF oscillation problems might render the amplifier
unstable even at no signal input. And this problem migth be very hard, not to say impossible,
to get rid of.

A simple way to test this is to apply a 10-20kHz square-wave (5-10Vp-p) as input signal in fig.1 while
measureing at the primary or TPC. If oscillations occurs, the OPT will, almost for certain,
have to be discarded.

See www.freeyellow.com/members5/rogerk/page11.html for more info/hints.